PARAMETER-FREE CLUSTER DETECTION IN SPATIAL DATABASES AND ITS APPLICATION TO TYPIFICATION
Karl–Heinrich Anders, Monika Sester · 2000
The automatic analysis of spatial data sets presumes to have techniques for interpretation and structure recognition. Such procedures are especially needed in GIS and digital cartography in order to automate the time-consuming data update and to generate multi-scale representations of the data. In order to infer higher level information from a more detailed data set, coherent, homogeneous structures in a data set have to be delineated. There are different approaches to tackle this problem, e.g. model based interpretation, rule based aggregation or clustering procedures. In the paper, a parameterfree graph-based clustering approach and an application in the domain of cartography, namely typification is presented. Typification is a generalization operation needed in order to present a set of objects by a subset of representatives. In this way, a collection of objects can be represented by fewer objects in a symbolic representation. An important prerequisite for the legibility of detailed representation is that the structure is preserved. This implies that object clusters are preserved.